[DO NOT MERGE] CI validation for oidc2 (llm-proxy on OIDC restructure)#568
[DO NOT MERGE] CI validation for oidc2 (llm-proxy on OIDC restructure)#568fabianvf wants to merge 15 commits into
Conversation
Signed-off-by: Jeff Ortel <jortel@redhat.com>
Signed-off-by: Jeff Ortel <jortel@redhat.com>
Signed-off-by: Jeff Ortel <jortel@redhat.com>
Signed-off-by: Jeff Ortel <jortel@redhat.com>
Signed-off-by: Jeff Ortel <jortel@redhat.com>
Signed-off-by: Jeff Ortel <jortel@redhat.com>
Signed-off-by: Jeff Ortel <jortel@redhat.com>
Signed-off-by: Jeff Ortel <jortel@redhat.com>
Signed-off-by: Jeff Ortel <jortel@redhat.com>
Signed-off-by: Jeff Ortel <jortel@redhat.com>
Signed-off-by: Jeff Ortel <jortel@redhat.com>
konveyor/tackle2-hub:latest is pre-OIDC (hub PR #1042 still open) and
crashloops on a Keycloak login. Per Jeff's note, the matching test
images for oidc2 are jortel/tackle2-{hub,ui}:oidc (both rebuilt
2026-05-30). Pin them so CI exercises the OIDC stack Jeff intends.
Do not merge -- real bundle must pin published konveyor images.
Signed-off-by: Fabian von Feilitzsch <fabian@fabianism.us>
|
Important Review skippedDraft detected. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
The new tackle.konveyor.io CRDs (IdpClient, IdentityProvider, LdapProvider) are reconciled by the operator. Add them to the Role's resources list so the operator's service account can manage them; without this the reconcile fails at the first IdpClient lookup with "User \"system:serviceaccount:.../tackle-operator\" cannot get resource \"idpclients\" ... Forbidden". Signed-off-by: Fabian von Feilitzsch <fabian@fabianism.us>
The OIDC UI redirects unauthenticated requests to a login page (302), so a curl Basic Auth call routed through the UI's /hub proxy never reaches the hub. Port-forward to service/tackle-hub instead and hit /services/llm-proxy/* directly, exercising the hub auth + RBAC + reverse proxy path we actually want to test. Also capture the HTTP status code on a failed chat-completion request (so auth/RBAC vs route/forward failures are distinguishable from an empty body) and dump tackle-hub logs in the debug step on failure. Signed-off-by: Fabian von Feilitzsch <fabian@fabianism.us>
…-proxy Mirror how real clients are expected to talk to the hub: present a short-lived credential (Basic admin:admin against the seeded user) to POST /auth/tokens, then use the returned PAT as a Bearer token for the /services/llm-proxy calls. The previous test sent Basic Auth on every chat-completion call, which works but isn't representative of the intended client flow (long-lived API key for service traffic). Verified locally against jortel/tackle2-hub:oidc: POST /auth/tokens returns 201 with a token; that token then authenticates the /services/llm-proxy/v1/chat/completions request. Signed-off-by: Fabian von Feilitzsch <fabian@fabianism.us>
|
Superseded by jortel#3 (clean PR against jortel:oidc2 with the test-only image pins dropped). End-to-end test goes green against jortel/tackle2-hub:oidc with Jeff's auth fix. |
|
Reopening to re-verify the updated test (UI-proxy happy path + hub-direct security check) against jortel/tackle2-hub:oidc. The actual fix is jortel#3 — this draft just exists for CI validation. |
Draft / do-not-merge. Scratch PR to run the real `llm-proxy-test` CI against Jeff's `oidc2` branch (his OIDC restructure, replaces the older PR #562 line of work).
Contains, on top of `oidc2`:
No other scaffolding has been added — running Jeff's restructured operator as-is to see what happens. Likely failure modes I'm watching for:
Refs: `oidc2` branch on Jeff's fork; jortel/tackle2-hub#1; konveyor/tackle2-hub#1042